Recognition of chlorophyll rings using YOLOv8

Area of Science:

• Marine biology
• Oceanography
• Artificial intelligence in environmental science

Background:

• Chlorophyll concentration is a key indicator of marine phytoplankton and primary productivity, vital for fisheries.
• Deep learning models are increasingly applied in ocean science for data analysis.

Purpose of the Study:

• To employ the YOLOv8 model for identifying elevated chlorophyll concentrations in chlorophyll rings around mesoscale ocean eddies.
• To statistically analyze the spatial distribution and generation of chlorophyll rings based on eddy characteristics.

Main Methods:

• Utilizing the YOLOv8 deep learning model for object detection of chlorophyll rings.
• Comparing YOLOv8 performance against Swin-Transformer and ResNet models.
• Statistical analysis of identified chlorophyll rings, considering eddy radii and life cycles.

Main Results:

• YOLOv8 demonstrated superior accuracy and generalization capability in identifying chlorophyll rings compared to other models.
• Statistical analysis revealed spatial distribution characteristics of chlorophyll rings.
• Quantified the generation of chlorophyll rings under different eddy radii and life cycles.

Conclusions:

• The YOLOv8 model provides a powerful tool for analyzing chlorophyll concentration data and understanding marine ecosystems.
• The study offers new insights into the dynamics and distribution of chlorophyll within marine environments.
• Findings contribute to improved marine fisheries management and ecological monitoring.